segformer-b0-finetuned-segments-sidewalk-oct-22

This model is a fine-tuned version of nvidia/mit-b0 on the julia-wenkmann/TennisSegmentation dataset. It achieves the following results on the evaluation set:

• Loss: 0.0577
• Mean Iou: 0.1977
• Mean Accuracy: 0.2635
• Overall Accuracy: 0.5273
• Accuracy Undefined: nan
• Accuracy Ball: 0.0
• Accuracy Playertop: 0.1196
• Accuracy Playerbottom: 0.6710
• Iou Undefined: 0.0
• Iou Ball: 0.0
• Iou Playertop: 0.1196
• Iou Playerbottom: 0.6710

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 6e-05
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 50

Training results

Training Loss	Epoch	Step	Validation Loss	Mean Iou	Mean Accuracy	Overall Accuracy	Accuracy Undefined	Accuracy Ball	Accuracy Playertop	Accuracy Playerbottom	Iou Undefined	Iou Ball	Iou Playertop	Iou Playerbottom
1.0528	1.67	20	1.1784	0.1882	0.2528	0.5396	nan	0.0	0.0471	0.7115	0.0	0.0	0.0411	0.7115
0.8484	3.33	40	0.8579	0.1435	0.1914	0.4261	nan	0.0	0.0	0.5741	0.0	0.0	0.0	0.5741
0.6695	5.0	60	0.6303	0.1414	0.1885	0.4196	nan	0.0	0.0	0.5654	0.0	0.0	0.0	0.5654
0.5808	6.67	80	0.5370	0.0901	0.1201	0.2674	nan	0.0	0.0	0.3603	0.0	0.0	0.0	0.3603
0.4415	8.33	100	0.4385	0.1263	0.1685	0.3751	nan	0.0	0.0	0.5054	0.0	0.0	0.0	0.5054
0.3955	10.0	120	0.3449	0.1177	0.1570	0.3496	nan	0.0	0.0	0.4710	0.0	0.0	0.0	0.4710
0.3597	11.67	140	0.3006	0.1186	0.1582	0.3521	nan	0.0	0.0	0.4745	0.0	0.0	0.0	0.4745
0.2761	13.33	160	0.2592	0.0968	0.1290	0.2873	nan	0.0	0.0	0.3871	0.0	0.0	0.0	0.3871
0.2343	15.0	180	0.2044	0.0987	0.1316	0.2930	nan	0.0	0.0	0.3948	0.0	0.0	0.0	0.3948
0.1953	16.67	200	0.1841	0.1258	0.1678	0.3736	nan	0.0	0.0	0.5033	0.0	0.0	0.0	0.5033
0.1676	18.33	220	0.1558	0.1394	0.1858	0.4137	nan	0.0	0.0	0.5574	0.0	0.0	0.0	0.5574
0.1486	20.0	240	0.1392	0.1448	0.1931	0.4299	nan	0.0	0.0	0.5792	0.0	0.0	0.0	0.5792
0.1281	21.67	260	0.1234	0.1497	0.1996	0.4445	nan	0.0	0.0	0.5989	0.0	0.0	0.0	0.5989
0.1138	23.33	280	0.1064	0.1407	0.1877	0.4178	nan	0.0	0.0	0.5630	0.0	0.0	0.0	0.5630
0.1031	25.0	300	0.0952	0.1495	0.1993	0.4438	nan	0.0	0.0	0.5979	0.0	0.0	0.0	0.5979
0.094	26.67	320	0.0896	0.1500	0.2000	0.4454	nan	0.0	0.0	0.6001	0.0	0.0	0.0	0.6001
0.0873	28.33	340	0.0889	0.1677	0.2236	0.4978	nan	0.0	0.0	0.6707	0.0	0.0	0.0	0.6707
0.0822	30.0	360	0.0772	0.1631	0.2175	0.4843	nan	0.0	0.0	0.6526	0.0	0.0	0.0	0.6526
0.0769	31.67	380	0.0739	0.1589	0.2119	0.4718	nan	0.0	0.0	0.6358	0.0	0.0	0.0	0.6358
0.0798	33.33	400	0.0694	0.1603	0.2137	0.4758	nan	0.0	0.0	0.6411	0.0	0.0	0.0	0.6411
0.0704	35.0	420	0.0654	0.1680	0.2241	0.4910	nan	0.0	0.0158	0.6564	0.0	0.0	0.0158	0.6564
0.0653	36.67	440	0.0633	0.1708	0.2278	0.4957	nan	0.0	0.0229	0.6604	0.0	0.0	0.0229	0.6604
0.0648	38.33	460	0.0610	0.1743	0.2324	0.5013	nan	0.0	0.0325	0.6648	0.0	0.0	0.0325	0.6648
0.0616	40.0	480	0.0598	0.1859	0.2479	0.5189	nan	0.0	0.0664	0.6773	0.0	0.0	0.0664	0.6773
0.0612	41.67	500	0.0586	0.1887	0.2517	0.5203	nan	0.0	0.0805	0.6745	0.0	0.0	0.0805	0.6745
0.0693	43.33	520	0.0584	0.1953	0.2604	0.5300	nan	0.0	0.1003	0.6810	0.0	0.0	0.1003	0.6810
0.0595	45.0	540	0.0567	0.1998	0.2664	0.5354	nan	0.0	0.1161	0.6831	0.0	0.0	0.1161	0.6831
0.0564	46.67	560	0.0556	0.2007	0.2676	0.5378	nan	0.0	0.1165	0.6862	0.0	0.0	0.1165	0.6862
0.0608	48.33	580	0.0555	0.2032	0.2710	0.5412	nan	0.0	0.1249	0.6880	0.0	0.0	0.1249	0.6880
0.0599	50.0	600	0.0577	0.1977	0.2635	0.5273	nan	0.0	0.1196	0.6710	0.0	0.0	0.1196	0.6710

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu121
• Datasets 2.17.1
• Tokenizers 0.15.2